Update README.md

README.md

@@ -1,3 +1,38 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+
+# MVInpainter
+[NeurIPS 2024] MVInpainter: Learning Multi-View Consistent Inpainting to Bridge 2D and 3D Editing
+
+[[arXiv]](https://arxiv.org/pdf/2408.08000) [[Project Page]](https://ewrfcas.github.io/MVInpainter/)
+
+Codes and models will be open-released soon!
+
+Codes organization is working in progress.
+
+### Dataset preparation (training)
+1. Downloading [Co3dv2](https://github.com/facebookresearch/co3d), [MVImgNet](https://github.com/GAP-LAB-CUHK-SZ/MVImgNet) for MVInpainter-O.
+Downloading [Real10k](https://google.github.io/realestate10k/download.html), [DL3DV](https://github.com/DL3DV-10K/Dataset), [Scannet++](https://kaldir.vc.in.tum.de/scannetpp) for MVInpainter-F.
+2. Downloading information of indices and masking formats from [Link](), and put them to ```./data```. Note that we remove some dirty samples from aforementioned datasets. We additionally provide complete foreground masks for MVImgNet through ```CarveKit```.
+
+### Pretrained weights (training)
+Put these weights to ```./check_points```
+1. [SD1.5-inpainting]()
+2. [AnimateDiff weights](). We revise the key name for easier ```peft``` usages.
+
+
+
+
+
+## Cite
+If you found our program helpful, please consider citing:
+
+```
+@article{cao2024mvinpainter,
+  title={MVInpainter: Learning Multi-View Consistent Inpainting to Bridge 2D and 3D Editing},
+  author={Cao, Chenjie and Yu, Chaohui and Fu, Yanwei and Wang, Fan and Xue, Xiangyang},
+  journal={arXiv preprint arXiv:2408.08000},
+  year={2024}
+}
+```